Electronic Supplementary Material (ESI) for Nanoscale Horizons.
This journal is © The Royal Society of Chemistry 2016
Supplementary Information
Alloy Oxidation as a Route to Chemically Active Nanocomposites of Gold
Atoms in a Reducible Oxide Matrix
P. Sutter,1,* S.A. Tenney,2 F. Ivars-Barcelo,2 L. Wu,3 Y. Zhu,3 and E. Sutter4
1Department
of Electrical and Computer Engineering, University of Nebraska-Lincoln, Lincoln, Nebraska
68588, USA; 2Center for Functional Nanomaterials, Brookhaven National Laboratory, Upton, New York
11973, USA; 3Condensed Matter Physics and Materials Science Department, Brookhaven National
Laboratory, Upton, New York 11973, USA; 4Department of Mechanical and Materials Engineering,
University of Nebraska-Lincoln, Lincoln, Nebraska 68588, USA
1. Morphology of AuSn on Ge support
Figure S1. Scanning electron microscopy images of a. Sn reference sample, prepared by
sputtering 6 nm Sn on Ge. b. AuSn sample, prepared by sequential deposition of Au and Sn, both
with an equivalent thickness of 6 nm. Scale bars: 100 nm.
2. Atomic-resolution HAADF-STEM of oxidized AuSn nanostructures
Figure S2. High-resolution high-angle annular dark field (HAADF; Z-contrast) image of oxidized
AuSn nanostructures, showing a bright crystalline core (high average atomic number, Z)
terminated by an amorphous oxide shell (darker; lower average Z). Outside the oxide shell, the C
support is shown dark (lowest Z). Scale bar: 5 nm.
Supplementary Information: “Alloy Oxidation as a Route to Chemically Active Nanocomposites
of Gold Atoms in a Reducible Oxide Matrix”
1
3. Chemical mapping of Sn and O in oxidized AuSn nanostructures by STEM EELS
Figure S3. Chemical mapping of Sn and O. a. Electron energy loss (EELS) spectra from the
acquired spectrum image. The spectra (1) (red) and (3) (green) are from the corresponding
rectangle areas in the high-angle annular dark field (HAADF) image (b), showing the dominant
features of the Sn-M4,5 edge and the O-K edge. The Au M4,5 signal (at 2206 eV) is not shown
here. The spectrum (2) (blue circles) is from the blue rectangle in (b), and shows a mixture of the
Sn-M4,5 and O-K edges. We determine the two components (Sn and O distribution) using multiple
linear least-squares fitting (MLLS) with a weighted linear combination of spectra (1) and (3). The
yellow line in (a) is the fitted spectrum with weight coefficients of 0.46 and 0.54 for spectrum (1)
and (3), respectively. (b) HAADF image. (c, d) The Sn and O distribution maps based on the
weighted coefficients. Scale bars in (b)-(d): 2 nm.
4. XPS data
Fig. S4 gives time-dependent XPS measurements of Sn 3d5/2 and Au 4f core levels of a
nanostructured AuSn alloy on Ge support. Sn 3d5/2 spectra of the AuSn alloy were
measured at different times during oxidation at room temperature in air, and compared
with the time evolution for a reference sample of pure Sn nanoparticles, prepared at the
same time and exposed to the same oxidation conditions. The Au 4f spectra for the AuSn
alloy were also measured over the same time intervals, and are compared with the Au 4f
spectrum of a bulk Au foil, used as a reference sample. The raw data (black symbols)
were deconvoluted by fitting to Gaussian-Lorentzian products after subtraction of a
Shirley-type background. For the Sn 3d5/2 spectra, dark and light blue shaded areas
indicate the fit components corresponding to oxidized Sn and metallic Sn0, respectively.
The Au 4f spectra could be fitted by single Gaussian-Lorentzians for the 4f5/2 and 4f7/2
peaks, which are shifted by a constant energy relative to the Au reference due to alloying
with Sn. No oxidized Au was detectable in the Au 4f spectra.
Supplementary Information: “Alloy Oxidation as a Route to Chemically Active Nanocomposites
of Gold Atoms in a Reducible Oxide Matrix”
2
Figure S4. a. Sn 3d5/2 XPS of a nanostructured AuSn alloy, exposed at room temperature to air
for different durations. Oxidation times (from bottom to top, see arrows): 1 week, 2 weeks, 3
weeks, 4 weeks. b. Same for a bulk Sn reference sample. For both AuSnox and Snox, the Sn 3d5/2
peaks can be deconvoluted into two components at binding energies of 485-485.5 eV (metallic
Sn0) and 487 eV (Sn oxide), respectively. In both cases, the Sn0 component decreases with
progressive exposure to air. c. Au 4f XPS of a nanostructured AuSn alloy, exposed at room
temperature to air for different durations. Oxidation times (from bottom to top, see arrow): 1 week,
2 weeks, 3 weeks, 4 weeks. Topmost panel: Au foil reference spectra.
5. Additional XPS spectra & quantification of the [O]:[Sn] ratio in the oxide shell
In addition to the Au 4f and Sn 3d photoelectron spectra, we also followed the O 1s core
level during the progressive oxidation of nanostructured AuSn alloys on Ge support. O1s
XPS spectra of AuSn after room temperature oxidation in air for 1 – 4 weeks are shown
in Figure S5a. The asymmetric peaks can be deconvoluted into two components, a lower
binding energy peak at 530.6 eV characteristic of the O 1s core level in Sn-oxides and a
peak at ~532 eV associated with Ge-oxides formed on the Ge support.
The ratio of the integrated XPS intensities of the Sn3d5/2 peak corresponding to Sn in the
oxide shell and of the O 1s component corresponding to Sn-oxide (shaded areas in Figure
S5b) was used to quantify the composition of the amorphous oxide shell of room
temperature oxidized AuSn nanostructures, in particular to determine the [O]:[Sn] atomic
ratio. Our previous work on the oxidation of pure Sn nanoparticles at room temperature
has shown the formation of an amorphous Sn(II) oxide (i.e., SnO) shell (E. Sutter et al.,
Part Part Syst Char 31, 879 (2014)), but the presence of Au in our AuSn samples could
modify the stoichiometry of the oxide shell. Our quantification shows that this is indeed
the case. Based on the integrated intensities of the relevant O 1s and Sn 3d5/2 components
Supplementary Information: “Alloy Oxidation as a Route to Chemically Active Nanocomposites
of Gold Atoms in a Reducible Oxide Matrix”
3
and taking into account the relative sensitivity factors (RSF = 0.66 for O 1s; RSF = 4.3
for Sn 3d5/2; C.D. Wagner et al., Surf. Interface Anal. 3, 211 (1981)), we find an [O]:[Sn]
atomic ratio of 1.3 for a sample oxidized for 4 weeks in air. This suggests that the
oxidation of AuSn at room temperature leads to the formation of non-stoichiometric
amorphous oxides comprising Sn, O, and Au.
Figure S5. a. Progression of the O1s XPS peak during oxidation of AuSn nanostructures on Ge
support at room temperature in air for 1 – 4 weeks. b. O1s and Sn3d core levels at t = 4 weeks.
The components used for quantification of the [O]:[Sn] atomic ratios are shaded light green (O1s)
and light blue (Sn3d5/2).
Finally, we note that the above quantification is insensitive to possible adsorbed O2 on
the surface of our samples. Previous XPS spectra by Kawabe et al. (Surf. Sci. 448, 101
(2000)) for O2 adsorbed on reduced (i.e., sub-stoichiometric) SnO2 at room temperature
showed an O1s peak centered at ~532.6 eV. Our measurements (Fig. S5) do not detect
significant spectral intensity at this binding energy, but are readily deconvoluted into only
two contributions: a broader peak from Ge-oxide and a narrower peak from Au-Sn oxide.
The latter, well-separated from the expected location of the O1s peak from adsorbed O2,
was used to quantify the composition of the amorphous oxide shells around Au-Sn
nanoparticles. Hence, our quantification would not be affected by contributions from
adsorbed O2.
6. Oxide thickness determination from XPS peak ratios
The thickness of the growing oxides (Fig. 2b of the paper) has been calculated from the
measured XPS peak ratios. The analysis was based on laboratory XPS data, obtained with
Al K excitation (h = 1486.6 eV). With this excitation, based on the universal curve the
inelastic mean free path of Sn 3d photoelectrons is  ~ 1.4 nm. Even though the Au-Sn
oxide is non-stoichiometric with a composition of ~SnO1.3 (see above), we used the mass
density of SnO (6.45 g·cm-3). The mass density of SnO2 (6.85 g/cm3) is only ~6% higher,
Supplementary Information: “Alloy Oxidation as a Route to Chemically Active Nanocomposites
of Gold Atoms in a Reducible Oxide Matrix”
4
i.e., using the density of SnO does not cause any errors within the statistical significance
of our data.
The oxide thickness on oxidized AuSn and -Sn samples was determined from the ratio
between the Sn 3d 5/2 intensities of Sn0 and Sn2+ (Fig. S4).
Oxide thickness determination for the -Sn reference sample: For -Sn we use the molar
mass MSn = 118.7 g·Mol-1 and density 7.365 g·cm-3; for SnO the molar mass MSnO is
134.7 g·Mol-1 and the density 6.45 g·cm-3. From these values, we find molar densities of
Sn (Sn = 0.062 Mol·cm-3) and of Sn in SnO (SnO = 0.048 Mol·cm-3).
We derive the thickness of the SnO like surface oxide on Sn nanostructures by integrating
the contributions of the oxide shell (thickness hox) and remaining metal core to the total
Sn 3d 5/2 XPS intensity, taking into account the inelastic mean free path, :
I(d, h) 
dh
 exp x  dx
d
To determine the oxide thickness, we used the following equation:
R  A(Sn 2 )  A(Sn 2 )  A(Sn 0 )  SnO I 0, hox  SnO I 0, hox    Sn I hox , r  .
Here, A(Sn2+) and A(Sn0) denote the areas (determined by peak fitting) of the 3d 5/2
peaks corresponding to Sn2+ and Sn0, respectively; (Sn) and (SnO) are the molar
densities of Sn in -Sn and in Sn(II) oxide; r is the thickness of the metallic core of the
nanostructures, determined from TEM to be (12 ± 2) nm; and I(d, h) is as defined above.
The oxide thickness, hox, was determined by numerically solving the above equation in
the software package Mathematica.
Oxide thickness determination for AuSn: In analyzing the thickness of the surface oxide
on AuSn, we used two limiting scenarios: (i) complete phase separation, i.e., formation
of a pure SnO shell and concentration of the Au in the metallic core; and (ii) totally
suppressed phase separation, i.e., formation of an oxide shell with unchanged atomic
ratio of Au and Sn.
The analysis in both cases follows that described above for Sn oxidation, but with
different values of the Sn molar densities:
(i) Complete phase separation:
 Sn density in SnO1.3: SnOx = 0.046 Mol·cm-3.
 Sn density in AuSn: AuSn = 0.053 Mol·cm-3.
(ii) Suppressed phase separation:
 Sn density in Au-SnO1.3: SnOx = 0.023 Mol·cm-3.
 Sn density in AuSn: AuSn = 0.053 Mol·cm-3.
The analysis shows a significantly reduced oxidation rate for AuSn than for the Sn
reference sample (see Fig. 2 of the paper).
7. Detection of Au in the growing surface oxide shell of AuSn nanostructures
Measurements of the time dependent intensity of the Au 4p 3/2 core level, which
provides a significantly reduced photoelectron escape depth and hence higher surface
Supplementary Information: “Alloy Oxidation as a Route to Chemically Active Nanocomposites
of Gold Atoms in a Reducible Oxide Matrix”
5
sensitivity compared to Au 4f, were used to detect if Au remains embedded in the
growing oxide and to quantify its amount. XPS measurements were compared with
simulations in two limiting cases: (i) complete phase separation, i.e., formation of a pure
SnO shell and enrichment of Au in the metallic core; and (ii) completely suppressed
phase separation, i.e., formation of a Au-Sn oxide shell with unchanged concentration
ratio of the two metals (here 1:1).
Our experiment followed the attenuation of the Au 4p 3/2 peak during oxidation of AuSn
over a period of 4 weeks. To simulate the data, we calculated the expected Au 4p 3/2
signal, based on the atomic densities Au in the two limiting cases. In case of complete
phase separation, a pure SnO shell would form, and as the shell grows thicker it
progressively attenuates the Au 4p intensity originating entirely in the metallic core. In
case of suppressed phase separation, Au 4p intensity originates in both the oxide shell
and the metal core, taking into account the different density of Au atoms in the oxide and
alloy.
To simulate the two scenarios, we integrated the contributions of the alloy core and oxide
shell to the overall Au 4p signal, assuming an inelastic mean free path ~ 1.4 nm
determined from the universal curve. The analysis used the formalism described in
section 2., with the following molar densities of Au in the different phases:
(i) Complete phase separation:
 Au density in SnO1.3: SnOx = 0.
 Au density in AuSn: AuSn = 0.053 Mol·cm-3.
(ii) Suppressed phase separation:
 Au density in Au-SnO1.3: SnOx = 0.023 Mol·cm-3.
 Au density in AuSn: AuSn = 0.053 Mol·cm-3.
The analysis summarized in Fig. 2c of the paper shows that phase separation is nearly
completely suppressed at the initial stages of oxidation. At a later stage the overall Au
content of the oxide decreases. However, a pure SnO shell is never achieved but instead
an oxide containing both Sn and Au is formed, as shown also in chemical mapping in
(S)TEM.
8. Mass spectroscopy analysis of near-ambient pressure CO-to-CO2 conversion
The reaction CO + ½ O2  CO2 was probed at room temperature by dosing different gas
mixtures (CO + O2) at near-ambient pressures into a ultrahigh vacuum chamber and
analyzing the reaction product (CO2) using a sampling capillary (0.75 mm bore) brought
into close vicinity (< 1 mm) of the surface. A constant partial pressure of O2 (0.1 Torr)
was maintained in the chamber while the CO partial pressure was varied in steps from ~0
(i.e., the CO background pressure in the UHV system) to 2 Torr. The partial pressures for
masses 28 AMU (CO), 32 AMU (O2), and 44 AMU (CO2) were followed by mass
spectrometry. Raw data for all measured sample composition are shown in Figure S5.
The data shown in the paper (Fig. 4) represent steady-state values of the CO to CO2
conversion efficiency, calculated from measurements at masses 28 AMU and 44 AMU
recorded after the gas pressures were stable to < 5% following each change in the CO
flow rate. The analysis assumes that all gas molecules that enter the sampling capillary
originate from the sample surface, i.e., are either molecules backscattered or desorbed (as
Supplementary Information: “Alloy Oxidation as a Route to Chemically Active Nanocomposites
of Gold Atoms in a Reducible Oxide Matrix”
6
reaction products or unreacted) from the sample. The CO to CO2 conversion efficiency 
is computed from the measured ratio of masses 44 AMU (CO2) and 28 AMU (CO and
CO2 fragments; assuming a negligible N2 background, which we verified independently)
via:
  I 44 AMU I 28AMU  0.11 I 44 AMU ,
where we have subtracted from the 28 AMU signal the contribution from cracked CO2
molecules, according to the cracking pattern of CO2:
(Mass/relative intensity): 44/100;
28/11; 16/9; 12/6; 45/1.
Figure S6. Mass spectrometer signals (masses 28, 32, and 44 atomic mass units, AMU) for
gases evolving from samples exposed 0.1 Torr O2 and different partial pressures of CO. The
gases were sampled by a capillary (0.75 mm inner diameter) brought to within 1 mm from the
sample surface. Different bimetallic Au-Sn alloys, oxidized at room temperature in air, were
probed: a. Au2Sn; b. AuSn; c. AuSn2; and d. AuSn4. e. Reference sample of oxidized Sn
nanostructures.
The gray shaded bars in Fig. 4 in the paper represent the as-calculated values of the
conversion efficiency for the different samples studied. Note that there is some scatter in
the oxygen pressure from sample to sample (see Fig. S6; nominal pO2 = 0.1 Torr; actual
pO2 ranging from 0.1 – 0.3 Torr), but this variability did not affect our conclusions
Supplementary Information: “Alloy Oxidation as a Route to Chemically Active Nanocomposites
of Gold Atoms in a Reducible Oxide Matrix”
7
because of the way the measurements were conducted (sampling at fixed O2 partial
pressure and at different CO partial pressures) and because the reported conversion
efficiencies were computed from actual (measured) partial pressures of the reactants. The
analysis of the CO to CO2 conversion efficiency described above assumes that the active
phase of our (Au-Sn)-oxide catalysts covers the entire sample surface. But this tends to
significantly underestimate the actual efficiency because the active catalyst, the dewetted
and then oxidized AuxSn1-x alloy, covers only a fraction of the surface of the Ge support.
This is illustrated in Fig. S7. In the left half of the TEM image shown in Fig. S7b the
inactive exposed Ge support is highlighted in red, whereas the active oxidized Au-Sn
alloy nanostructures are shown in gray.
From Fig. S7b, the actual surface area covered by the active (Au-Sn)-oxide catalyst is
estimated to be ~70% of the total sample area. We have taken this into account by
multiplying the measured efficiency values by an appropriate correction factor (1.43).
The corrected values are given as orange colored bars in Fig. 4 of the paper. For
completeness, we are reporting in Fig. 4 both the uncorrected (gray) and corrected
(orange) values, as illustrated in Figure S7a.
Figure S7. Geometric correction of the measured CO to CO2 conversion efficiency to account for
to the partial surface coverage of the active oxidized Au-Sn alloy phase. a. Illustration of the data
representation in fig. 4 of the manuscript. b. TEM image illustrating the areas covered by the
active phase and by the exposed Ge support (shaded red on the left hand side of the
micrograph).
9. Reversible CO adsorption in low-energy ion scattering spectroscopy (LEIS)
To further support the suggested CO oxidation reaction pathway – a mechanism in which
embedded Au atoms provide active low-coordinated Au sites for adsorption of CO and
reaction to CO2 with oxygen from the surrounding oxide surface – we performed LEIS
measurements during CO exposure of an oxidized AuSn sample. These measurements are
summarized in Fig. S8.
Supplementary Information: “Alloy Oxidation as a Route to Chemically Active Nanocomposites
of Gold Atoms in a Reducible Oxide Matrix”
8
Figure S8. LEIS during reversible CO adsorption. a. LEIS signal from Au during CO dosing
(pressure: 210-7 Torr, sample temperature: 120 K). Progressive reduction of the Au signal due to
shadowing of Au sites by adsorbed CO (inset). Right: Cumulative CO dose. b. Recovery of the
Au LEIS signal during thermal desorption of CO in vacuum, shown on the same scale as the data
in a. Right: Sample temperatures. Inset: oxygen LEIS spectra at 120 K before CO adsorption
(green), after 33.6 L CO dose (blue), and after desorption of CO by heating to 297 K (red). The
increase in the O LEIS signal is consistent with C binding to Au (i.e., exposure of O) during CO
adsorption.
LEIS measurements in CO were carried out at minimal total He+ dose on the sample to
suppress the modification by sputtering by the low-energy ions. Adsorption of CO at a
sample temperature of 120 K causes a progressive decrease in the LEIS signal from Au,
consistent with the shadowing of Au by adsorbed CO molecules (Figure S8a). The
reversibility of the adsorption is shown by LEIS measurements during heating of the
sample from 120 K to room temperature (Figure S8sb). In this process, the intensity of
the Au LEIS signal recovers to >97% of the original intensity prior to CO adsorption,
demonstrating the nearly complete desorption of CO from the Au sites close to room
temperature. Comparing the oxygen LEIS signal before and after CO adsorption, as well
as after thermal desorption, we observe a distinct increase in the intensity upon CO
adsorption due to the binding of CO in a way (C down) that projects the O atom toward
vacuum. As with the Au LEIS signal, this change is completely reversible upon thermal
desorption of CO from the surface. Our observations are consistent with molecular
adsorption of CO at Au surface sites, and relatively weak binding with binding energy of
the order of a few kBT at room temperature.
Supplementary Information: “Alloy Oxidation as a Route to Chemically Active Nanocomposites
of Gold Atoms in a Reducible Oxide Matrix”
9